Lithium-ion battery health estimate based on electrochemical impedance spectroscopy and CNN-BiLSTM-Attention

Abstract

To ensure the safe operation and optimal performance of lithium battery systems, accurately determining the state of health (SOH) of the batteries is crucial. Research over the past few decades has shown that techniques based on electrochemical impedance spectroscopy (EIS) offer some advantages over traditional methods relying on voltage, current, and temperature. In this paper, we propose a novel approach for assessing the SOH of lithium-ion batteries using a CNN-BiLSTM-Attention model. By combining the effectiveness of bidirectional long short-term memory (BiLSTM) neural networks, known for their efficiency in long sequence prediction, with convolutional neural networks (CNN) capable of automatically extracting EIS features, we create a unique CNN-BiLSTM model. Additionally, an attention mechanism is incorporated to enhance the model’s accuracy and processing speed. This approach enables faster and more effective feature extraction while minimizing information loss from historical data. Experimental results demonstrate that the proposed model achieves higher estimation accuracy compared to other popular data-driven methods. When compared to the benchmark BiLSTM and CNN-BiLSTM models, the AC-BiLSTM model reduces the root mean squared error (RMSE) by 93.9% and 71.4%, respectively. These findings highlight the significant practical value of the proposed approach.